IEEE P

roof

IEEE JOURNAL OF RADIO FREQUENCY IDENTIFICATION 1

Constrained Safety-Integrity Performance ofThrough-the-Arms UHF-RFID Transcutaneous

Wireless Communication for theControl of Prostheses

Carolina Miozzi , Giovanni Saggio , Emanuele Gruppioni , and Gaetano Marrocco

Abstract—Advanced prostheses for recovery of arm amputa-1

tion can be nowadays controlled by the electromyographic (EMG)2

signals. Implanted myoelectric sensors, suitable to transcutaneous3

wireless reading, permit to improve the signal-to-noise ratio. This4

paper explores the feasibility of a through-the-arm telemetry5

link based on the radiofrequency identification in the UHF band6

(860–960 MHz). The proposed model accounts for the power7

sensitivity of the commercial devices, the constraints enforced8

by the exposure regulations (SAR) and by the communication9

integrity (BER). The reliability of the link is evaluated against10

possible misalignments between sensors and the reading unit.11

Results demonstrate that the transcutaneous link can be in some12

case limited by integrity constraints but can be nevertheless cor-13

rectly established by means of less than 23 dBm input power14

(full compatible with embedded readers). The link is moreover15

robust against angular displacement up to at least �φ = ±35 ◦16

and linear displacement up to 2.5 cm.17

Index Terms—Radiofrequency identification, robotic limb,18

transcutaneous wireless communication, implantable device.19

I. INTRODUCTION20

STATE of the art prostheses for arm amputation recov-21

ery are controlled by the electromyographic (EMG) sig-22

nals [1] generated by the electric potential difference (typically23

70-90 mV peak [2]) between the outside and the inside of the24

muscular cells during muscle contraction. EMG signals are25

generally collected by surface-mounted sensors [3], [4] that26

have however several disadvantages since they are exposed27

to the variation of skin impedance due to sweat and could28

detach. They are moreover characterized by a low signal-to-29

noise ratio due to the absorption of the myoelectric signals by30

muscle/fat/skin layers. To overcome some of the above prob-31

lems, myoelectric sensors could be implanted in the arms [1],32

[3] and be interrogated from the prosthesis through a transcuta-33

neous wireless link. A pioneering system [5], [6] operating in34

Manuscript received January 21, 2019; revised April 27, 2019; acceptedMay 20, 2019.AQ1 This work was supported in part by the Istituto Italiano diTecnologia and in part by the Istituto Nazionale Assicurazione Infortuni sulLavoro—Centro Protesi under Contract PPR AS 1/1, Task 1.5. (Correspondingauthor: Gaetano Marrocco.)

C. Miozzi, G. Saggio, and G. Marrocco areAQ2 with the University of RomaTor Vergata, 00133 Rome, Italy (e-mail: gaetano.marrocco@uniroma2.it).

E. Gruppioni is withAQ3 INAIL Centro Protesi, Research and Training Area,Budrio, Italy.

Digital Object Identifier 10.1109/JRFID.2019.2921097

battery-less mode comprised two coupled planar coils (diam- 35

eters: 6 mm for implanted antenna and 2 cm for external unit 36

placed into the socket) tuned at 8 MHz. Imperfect mounting 37

and dismantling of the prosthesis (at least once per day) may 38

in this case produce misalignment [7] between sensors and 39

interrogator and accordingly a degradation of the link. The 40

Implantable MyoElectric Sensor (IMES) [8], [9], [10] (diam- 41

eter: 2.5 mm; length: 16 mm) resorted to a 121 kHz link. The 42

interrogator device was in this case a large coil wrapped onto 43

the socket that hence needs a manual customization for each 44

patient. 45

This paper explores the feasibility of a different kind 46

of wireless link involving the Radiofrequency Identification 47

(RFID) technology in the UHF band (860-960 MHz) that is 48

expected to offer some potential advantages such as increas- 49

ingly availability of off-the-shelf sensor-oriented microchip 50

transponders and readers and more freedom in the shaping 51

of the interrogating antenna. This, could be attached onto the 52

prosthesis like a thin plaster, reducing the need of manual cus- 53

tomization. Moreover, as a UHF reader allows a much longer 54

read range than lower-frequency ones, it could also enable the 55

prosthesis to interact, with the external environment according 56

to the framework of Internet of Things and Smart Spaces [11], 57

thus providing the patient with augmented senses [12]. UHF 58

systems have however to face within the much higher losses 59

of the human body. The expected consequence is a stronger 60

power absorption inside tissues so that the constraint over 61

the maximum allowed Specific Absorption Rate (SAR) could 62

become critical for the feasibility of the transcutaneous com- 63

munication based on backscattering modulation. 64

Preliminary tradeoff analysis for the RFID link of implanted 65

RFID sensors, including also SAR considerations, were 66

reported in [13] and in [14] concerning the monitoring of 67

vascular stents and orthopedic prostheses, respectively. But 68

in those cases the interrogating antenna was placed up to 69

20-30 cm from the skin and the estimated SAR resulted 70

greatly below the limits. The feasibility of transcutaneous 71

UHF-RFID telemetry, with the interrogating antenna placed 72

this time at a very close distance from the skin, was inves- 73

tigated in [15] concerning wireless brain-machine interfaces. 74

SAR compliance revealed a potential limiting factor to estab- 75

lish the link. Nevertheless, such analysis only considered 76

the direct link (from reader to the tag) while no result was 77

2469-7281 c© 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

IEEE P

roof

2 IEEE JOURNAL OF RADIO FREQUENCY IDENTIFICATION

Fig. 1. Sketch of transcutaneous communication system for the control ofupper-limb myoelectric prostheses.

provided concerning the quality of the data-link. Overall, the78

state of the art does not provide enough information to assess79

the feasibility of an RFID-UHF transcutaneous link in real80

conditions accounting for all the electrical and geometrical81

parameters.82

Starting from the above results, and from the prelimi-83

nary conference paper [16] with over-simplified models, this84

work introduces a more detailed electromagnetic formulation85

(Section II) and arm representation. It exploits both the wire-86

less power transfer link and the backscattering communication,87

realistic sensitivities of sensor-oriented microchip and an ad-88

hoc formulation of the Bit Error Rate (BER) for a two-port89

RFID system (Section III). The goal is to estimate through90

numerical analysis (Section IV) and some preliminary exper-91

imental corroboration (Section V), read volume of implanted92

sensors inside the arm that is compatible with all the electrical93

and safety parameters [17], also including possible misalign-94

ment of the sensors and the communication integrity, and95

to compare them with the performance of corresponding HF96

systems.97

II. THROUGH-THE-ARM TELEMETRY SYSTEM98

A typical upper limb myoelectric prosthesis (Fig. 1) com-99

prises an electro-mechanical device replacing the hand itself100

with grasp capability and a custom-made resin socket, suitable101

to perfectly adhere to the arm stump of the patient, where the102

prosthesis will be mechanically inserted.103

EMG sensors generally comprise a couplet of conductive104

electrodes to be applied at a few millimeters distance [18] and105

require a 1-5 V bias voltage. The myoelectric signal detected106

from the electrodes is inversely proportional to the distance107

from the source [2], but after an internal amplification to108

the EMG sensor, the output signal peak is of the order of109

0.5-2 V [19], depending on the gain. Typical sampling rate110

for the accurate signal tracing is of the order of 1 kHz [18].111

However, a much coarser sampling frequency could be enough112

to just extract a few features of the signal to control some113

gestures of the hand prosthesis [20]. The exact determination114

of the minimum sampling rate in case of implanted sensors,115

Fig. 2. Multilayered forearm and socket model with indication of theinterrogation antenna and of four possible positions of the implanted loop.Length: 25 cm; cross section: 70 mm × 50 mm. Electromagnetic parameters:skin: εr = 41.6, σ = 0.8 S/m; fat: εr = 5.5, σ = 0.05 S/m; muscle:εr = 55.1, σ = 0.9 S/m; bone: εr = 12.5, σ = 0.1 S/m; epoxy resin: εr = 4,σ = 6 · 10−4 S/m.

when a better signal to noise ratio is expected, is processing- 116

dependent and is currently outside the scope of this paper that 117

is instead only focused on the radiofrequency transcutaneous 118

link independently on the EMG sensor features ad shape. 119

In order to implement a wireless sampling of the myoelec- 120

tric signals directly at the level of the muscular fibers of the 121

stump, it is assumed to implant small-size RFID transpon- 122

ders between the fat and the muscle layers at positions {A, B, 123

C, D}, where EMG sensors should be ideally located (Fig. 2). 124

Positions A and C are hereafter considered as the reference 125

implants to sample both the agonist and antagonist forearm 126

muscles that are involved in the main movements of fin- 127

gers and wrist. Position B and D refer to sensors implanted 128

in the lateral fat-muscle interface w.r.t. the bone, and they 129

are also representative of extreme cases of axial migration 130

of implanted tags. Sensors will be powered and interrogated 131

by means of one or more low-size and conformable anten- 132

nas placed on the interface between the prosthesis and the 133

socket. The wireless power transfer and the interrogation of 134

the implanted transponders by the reader are based on the full- 135

duplex Radiofrequency Identification standards EPC-Gen-2 so 136

that transponders will send back the collected data through a 137

backscattering modulation of the interrogating field [21]. 138

As the implanted transponders do not possess an 139

autonomous source, the establishment of a robust commu- 140

nication link is conditioned to the power that a small-size 141

battery-driven UHF reader is allowed to radiate in order to 142

overtake the power losses of the tissues and activate the 143

implanted sensors. Such a power is moreover constrained to 144

IEEE P

roof

MIOZZI et al.: CONSTRAINED SAFETY-INTEGRITY PERFORMANCE OF THROUGH-THE-ARMS UHF-RFID TRANSCUTANEOUS WIRELESS COMMUNICATION 3

Fig. 3. Antennas involved in the evaluation of the transcutaneous teleme-try system and their tuning capability: (a) Microstrip slot as interrogationantenna and its reflection coefficient (L = 40 mm, W = 25 mm, LP = 19.5 mm,d = 6 mm and s = 3 mm) and (b) the implanted rectangular loop and itspower transfer coefficient (LA = 12.6 mm, LB = 7.2 mm and s = 2.5 mm).

comply with regulation limits concerning the exposure of the145

human body to electromagnetic fields.146

A. Arm Model147

The arm stump is simulated by a realistic multi-layered148

cylindrical model derived from the extrusion of a cryo-149

sectioned forearm picture [22] (Fig. 2). The arm is inserted150

into a coaxial cylinder emulating the prosthetic socket, made151

of epoxy resin and partly leveled on top to easily allocate152

the slot antenna. All the electromagnetic computations to be153

showed next were performed by CST Microwave Studio 2017,154

Time-Domain Solver.155

B. Interrogation Antenna and Implanted Transponder156

The considered antennas are based on layouts already used157

in biomedical applications, such as a small loop acting as158

implanted tag and a microstrip slot working as interrogator.159

The microstrip slot is derived from a typical applicator for160

microwave Hyperthermia (also known as current sheet [23])161

and later-on used for RFID sensing in [24]. The antenna162

(Fig. 3a) is made of two patches facing each other and shorted163

to the ground plane at opposite edges by means of two verti-164

cal stripes. The substrate is a 3 mm layer of Closed-cell PVC165

foamboard (εr = 1.55, σ = 6 ·10−4 S/m). The antenna can be166

adjusted around 867 MHz by varying the length of the stripes167

(parameter d).168

The rectangular loop geometry is derived from a commer-169

cial general-purpose miniaturized tag (Mecstar Loopetto [25])170

at the purpose to simplify the experimental evaluation as171

Fig. 4. Simulated SAR delivered by the microstrip slot inside the layeredmodel of the arm’s stump corresponding to Pav,G = 1 W and averaged onto10 g of tissue.

described later-on. The tag is assumed to be made of an 172

aluminum trace (thickness: 10 μm) deposited onto a PET sub- 173

strate (εr = 3, σ = 1.6·10−2 S/m, thickness: 10 μm) (Fig. 3b). 174

The width of the smaller sides can be varied in order to tune 175

the antenna to the chip impedance ZC = (12 − j120)� at 176

867 MHz and perform some manual adjustment later on in 177

the experimentations. 178

C. Specific Absorption Rate 179

Fig. 4 shows the computed SAR delivered by the trans- 180

mitter inside the arm stump (without the implanted antenna) 181

corresponding to a unitary input available power Pav,R at the 182

reader side. Values are much higher than in the case of a 183

remote interrogating antenna as found in [14]. By considering 184

that the safety regulation [26] requires the maximum allowed 185

SAR in the arms, averaged onto 10 g, to be less than 4 W/kg, 186

it follows by linearity that the maximum power emitted by the 187

reader must be Pav,R < PSARav,R � 23 dBm. 188

III. LINK PARAMETERIZATION AND CONSTRAINTS 189

The above transcutaneous link is properly modeled [27] 190

by a two-ports network and by its self Zii and mutual Zij 191

impedances. Port 1 and port 2 refer to the terminals of the 192

interrogation antenna and of the implanted tag, respectively. 193

ZG and ZL are the impedance of the generator connected to 194

the reader and the RF-equivalent chip impedance. This rep- 195

resentation accounts for possible impedance mismatch at the 196

reader-antenna port as well as at the interconnection between 197

the implanted antenna and the chip, caused by the disturbing 198

effects of human body [12] and by the mutual coupling among 199

the two antennas. 200

A. System Gains 201

Forward and backward links are parameterized [14] by 202

means of the Transducer Power Gain GT and Round-Trip 203

Power Gain GRT , respectively, defined [12] as: 204

GT = PR→T

Pav,R= 4RchipRG|Z21|2

∣∣(Z11 + ZG)

(

Z22 + Zchip)− Z12Z21

∣∣2

(1) 205

GRT = PR←T

Pav,R= 1

4

∣∣∣Γin

(

ZON)

− Γin

(

ZOFF)∣∣∣

2(2) 206

IEEE P

roof

4 IEEE JOURNAL OF RADIO FREQUENCY IDENTIFICATION

where �in(Zmod) = [Zin(Zmod) − Z0]/[Zin(Zmod) + Z0] is the207

reflection coefficient at the input port of the network corre-208

sponding to the two impedance states Zmod = {ZON, ZOFF}209

of the microchip during the backscattering modulation. Zin =210

Z11 − Z12Z21/(Z22 + Zmod) is the input impedance seen by211

the reader toward the networks and Z0 = ZG = 50 � is the212

equivalent input impedance of the reader’s receiver after the213

circulator [14].214

Hence, by assuming ZON = Zchip and |ZOFF| � |ZON |, and215

introducing Zin and �in in (2) the round-trip gain can be finally216

expressed by simple mathematical manipulations as:217

GRT =∣∣∣∣∣

ZGZ212

Z11 + ZG

∣∣∣∣∣

1∣∣(Z11 + ZG)

(

Z22 + Zchip)− Z12Z21

∣∣2. (3)218

B. Bit Error Rate219

The quality of the whole link is quantified by the BER that220

provides an indication of the amount of lost bits during the221

communication between the reader and the implanted tag and222

hence is particularly meaningful for the measurement of phys-223

iologic signals. For an RFID link, the BER can be expressed224

as [28] BER = 12 erfc( |V0|m

2√

2σ), where V0 = 2

√

2RGPav,R is225

the input voltage at the reader and σ is the standard devia-226

tion of an Additive White Gaussian Noise (AWGN) corrupting227

the received signal. The modulation index m = |Γin(ZON) −228

Γin(ZOFF)|/2, can be hence derived from (2) as:229

m = √

GRT (4)230

For the sake of the simplicity, the noise is referred to the231

only body plus the reader antenna at a same temperature T so232

that σ = s√

KTRG�f [29], where �f is the frequency band233

of the RFID link (20 MHz at most) and K the Boltzmann234

constant. Then:235

BER = 1

2erfc

(

GRT

4

√

Pav,R

KT�f

)

. (5)236

Typical UHF RFID readers require BER < 10−3 [21]237

so that, by conservatively assuming �f = 20 MHz and238

T = 310.15 ◦K (human body temperature: 37◦C), the min-239

imum acceptable combination of round-trip gain and input240

power is (in decibel notation):241

GRT + 1

2Pav,R ≥ −56 dB. (6)242

For instance, by assuming Pav,R = 23 dBm the minimum243

value of the round-trip gain that permits to establish a reliable244

link is GRT > −52.5 dB.245

C. Power Margin for Reliable Links246

A reliable communication link can be established by pro-247

viding that the power emitted by the reader Pav,R is simultane-248

ously such i) to deliver enough power to activate the implanted249

microchip transponder, ii) to generate a backscattered signal250

that is strongly enough to be detected by the reader’s receiver251

with a proper BER as in (6) and, finally, iii) to be compliant252

with SAR limits inside the body. At this purpose, the power253

PR→T delivered by the reader to the tag’s chip must exceed254

the chip sensitivity pC (Forward Link). Simultaneously, the 255

backscattered power PR←T from the tag toward the reader’s 256

antenna, that is collected by the receiver, has to exceed the 257

power sensitivity pR of the reader (Backward Link). In formu- 258

las, it is easy to show that the available power threshold at the 259

reader to activate the tag is (parameters expressed in dB): 260

Pav,R > PR→Tav,R |dB = pC − GT . (7) 261

The power threshold, so that reader’s receiver is allowed 262

to recognize and decode the response of the tag, is instead 263

(parameters in dB): 264

Pav,R > PR←Tav,R |dB = pR − GRT . (8) 265

An additional constraint on the emitted power comes from 266

the BER level (6), so that: 267

Pav,R > PBERav,R|dB = −112 dB− 2GRT . (9) 268

It is worth noticing that, in case the round trip gain is such 269

that 270

GRT < −112 dB− pR, (10) 271

the backward link is limited by the BER constraint indepen- 272

dently on the input power. Overall, the minimum reader power 273

to enable both the forward and backward links and provide a 274

reliable BER links is hence: 275

Pminav,R|dB = max{pC − GT , pR − GRT ,−112 dB− 2GRT} (11) 276

Finally, by denoting with Pmaxav,R the maximum available 277

power the reader is capable to emit (generally 0.25-1 W), 278

the following constrained power margin M of the link is 279

introduced: 280

M(pC, pR, GR, GRT) = min{

Pmaxav,R, PSAR

av,R

}

− Pminav,R − P0 (12) 281

with P0 a safe value to account for non fully controllable 282

parameters of the system. The communication can be therefore 283

established and reliable for a specific implanted transponder 284

if M>0. 285

IV. ESTIMATION OF READ REGION IN THE ARM AND 286

TOLERANCE TO MISALIGNMENTS 287

This Section investigates the two-ports gains, and accord- 288

ingly the overall power margin, that can be achieved for 289

the above described arrangement of the reader and the tag 290

implanted as in Fig. 2, also considering some possible orienta- 291

tions of the tags. Then, for the most appropriate configurations, 292

the effect of a partial misalignment between tag and reader 293

is analyzed in order to quantify the resilience of the UHF 294

transcutaneous link against the possible variability of the 295

prosthesis-stump mounting. 296

A. Constrained Power Margin Versus the Tag Position 297

A first set of simulations refers to the tag placed in the posi- 298

tion {A, B, C} (position D being equivalent to the position B) 299

for possible orientations of both reader and tags as sketched 300

in Table I. The arrows indicate a reference orientation of the 301

two antennas with respect to the reference system in Fig. 2. 302

IEEE P

roof

MIOZZI et al.: CONSTRAINED SAFETY-INTEGRITY PERFORMANCE OF THROUGH-THE-ARMS UHF-RFID TRANSCUTANEOUS WIRELESS COMMUNICATION 5

Fig. 5. Power gains of the two-ports network that parameterize the near-fieldreader-sensor interactions for a loop placed at position A(x) (just underneaththe slot antenna).

TABLE ITRANSDUCER AND ROUND-TRIP POWER GAINS (IN DB) AT 867 MHz FOR

DIFFERENT COMBINATIONS OF THE MUTUAL ORIENTATION READER-TAG

For instance, the notation A(x) refers to a loop placed at the303

position A and oriented so that the arrow is parallel to the x304

axis.305

An example of frequency profile of the link gains is reported306

in Fig. 5 for tag in position A(x). Both the transducer and the307

round-trip gains are rather stable in the whole RFID UHF band308

due to the intrinsic broad-banding effect of the lossy tissue of309

the arm model.310

Results in Table I, corresponding to data at the UHF-RFID311

European frequency f = 867 MHz, show that the most efficient312

mutual orientations among the interrogating slot antenna and313

implanted loops are such that their corresponding arrows in314

Table I are mutually orthogonal. In particular, a reader antenna315

aligned orthogonal to the arm axis (labeled as “Reader (z)”)316

looks as the preferred arrangement.317

To give some meaning to numbers of Table I, it is worth318

recalling that typical values of two-ports gains for implanted319

antennas into the limbs [14] (elbow, shoulder, hip knee)320

in case of an external reader (at 90cm from the skin) are321

{GT , GRT} � {−45 dB,−90 dB}, while the transducer gain322

for a sensor placed onto the fingertip and a wrist reader was323

found [12] to be GT � −45 dB.324

For the most efficient A, B and C configurations (indi-325

cated in boldface in Table I), it is hence possible to estimate326

from (7), (8) the threshold powers of the forward and back-327

ward links as well as the overall constrained margin in (12).328

For this purpose it is assumed a typical power sensitivity of the329

reader (pR = −60 dBm) and two sensitivities of state of the art330

sensor-oriented microchips {pC1 = −5 dBm, pC2 = −10 dBm}331

as in the AMS-SL900A [30] and RFMicron Xerxes [31] or332

FARSENS Rocky 100 [32], respectively. Following a standard333

RFID interrogation, they are moreover capable to provide an334

TABLE IITHRESHOLD POWERS IN [DBM] FOR THE FORWARD AND THE BACKWARD

LINKS AND THE MAXIMUM BER AND CONSTRAINED POWER MARGIN

[IN DB] FOR Pmaxav,R = 30 dBm AND P0 = 3 dB

output voltage that is compatible with the EMG biasing so 335

that local battery can be avoided. Finally, the maximum power 336

that can be emitted by an embedded reader is assumed to be 337

Pmaxav,R = 30 dBm and the safeguard margin P0 = 3 dB from our 338

experience. Results summarized in Table II show that the link 339

is almost always forward-limited since PR→Tav,R > {PR←T

av,R , PBERav,R} 340

and the constrained margins are positive for the configurations 341

A and C. The transcutaneous link is hence feasible for sensors 342

implanted on both the agonist (A) and antagonist (C) muscles 343

even in case of the low-sensitivity microchip (M = 0.9 dB 344

in the worst case). Instead, sensor placed in B results always 345

un-readable (M = −20 dBm), due to BER constraint which 346

demands for a too high input power even in case the bet- 347

ter chip sensitivity was used. Anyway, such a sensor could 348

be interrogated, if needed, by a second slot antenna placed 349

at orthogonal position w.r.t. the reference one (results labeled 350

as B’). 351

It is however worth noticing the interesting result that 352

the BER constraint always dominates the backward link as 353

PBERav,R > PR←T

av,R . Moreover, for the case of the high sen- 354

sitivity microchip (pC = −10 dBm) and sensor placed at 355

the deepest distance (case C), the BER constraints dominate 356

the whole link, being the bottleneck of the communication 357

(PBERav,R > PR→T

av,R ). 358

B. Tolerance to Misalignments 359

Fig. 6a-b show the degradation of the power margin when 360

the slot antenna is incrementally shifted along the z direc- 361

tion (longitudinal misalignment) and along the azimuthal angle 362

φ = 0 − 90 ◦ (axial misalignment) on the same cross-section 363

(z = const). The system looks rather robust to misalignment 364

(M > 0) up to �z = 1cm (C)−2.5 cm (A), practically indepen- 365

dently on chip sensitivity, and up to �φ = 30◦(A) − 40◦(C) 366

even in the worst cases. The effect of misalignment could be 367

mitigated, as discussed in [33], by increasing the size of an on- 368

body interrogating antenna at the price of a further reduction 369

of power efficiency of the system, so that a proper trade-off 370

must be found. As expected, the margin of the tag in position C 371

is narrower than for tag in position A. 372

C. Comparison With Low-Frequency Transcutaneous Links 373

The achieved performance of the UHF RFID link are com- 374

pared in Table III with those of some low-frequency inductive 375

transcutaneous links that have been investigated for similar 376

IEEE P

roof

6 IEEE JOURNAL OF RADIO FREQUENCY IDENTIFICATION

Fig. 6. Constrained power margin to establish the RFID link between thecontacting slot antenna and the implanted tags at positions A (continuouslines) and C (dashed lines) for (a) longitudinal and (b) azimuthal shift of thereader’s with respect to the perfectly aligned arrangement. Margin evaluatedas in (12) for Pav,G = 23 dBm, P0 = 3 dB and the two considered chipsensitivities pC = −5 dBm (black lines) and pC = −10 dBm (gray lines).

TABLE IIICOMPARISON AMONG PERFORMANCE OF LOW FREQUENCY AND

UHF TRANSCUTANEOUS LINKS

applications. Since the size of the devices are not fully equiv-377

alent, such a comparison has to be therefore considered as378

qualitative. Overall, in spite of the much higher losses of the379

human tissues in the UHF band, the link performance are380

rather similar. The proposed architecture looks slightly more381

insensitive to longitudinal displacement but less efficient in the382

power transfer, as expected.383

V. PROTOTYPES AND EXPERIMENTAL EVALUATION384

To corroborate the above numerical analysis, the stump of385

the forearm was emulated by minced meat inserted into a real386

prosthetic socket. Due to the absence of lower loss tissues,387

like bones and fat, this configuration will provide more con-388

servative results (lower gains) than in the previous numerical389

simulations.390

The prototypes of both interrogating and implanted antennas391

are shown in Fig. 7. Tag was coated with an ultra-thin biocom-392

patible film (Rollflex film from Master-AID, 22 μm thickness).393

It is worth clarifying that the considered rectangular loop tag394

includes the Impinj Monza R6 [34] UHF RFID microchip that395

does not support sensing capability as its low power thresh-396

old pC = −20 dBm makes experimentations much easier than397

in case of sensor-oriented chips. The two-port gains, that are398

independent on the chip sensitivity, can be therefore derived399

even for the most challenging configurations, with the benefit400

1As the information about the link deterioration due to angular misalign-ment can not be directly identified in the referenced papers, such a degradationis roughly assumed to be proportional to |cos�φ|2 where �φ means the anglebetween the normal axes of transmitting and receiving coils.

Fig. 7. Manufactured prototypes of (a) the interrogating slot antenna and(b) the loop coated by bio-compatible 22 μm polyurethane film.

Fig. 8. (a) Experimental phantom made by minced meat filling a realprosthetic socket: Detail of the subcutaneous insertion of the tag (top) andplacement of the interrogating antenna over the external surface of the pros-thesis (bottom). (b) Measured reflection coefficient of the reader’s antennawhen it was placed over the real prosthetic socket filled with minced meat.

to can be applicable also to future generations of low-power 401

sensing-oriented chips. The relevance of obtained results will 402

be however discussed later on also for lower sensitivity chips. 403

The slot antenna was manufactured with adhesive copper 404

and its return loss (for placement over the socket) shows that, 405

apart for a 20 MHz shift due to the difference between the 406

simulated ideal model and the forearm phantom (Fig. 8b), the 407

11.5% (at −10 dB) bandwidth looks adequate for a robust 408

interrogation of the tag in the UHF RFID band. 409

A. Measurements 410

The tag was implanted subcutaneously, such as in Fig. 8a. 411

The measurements setup comprised an interrogating broad- 412

band antenna connected to the Voyantic Tagformance station. 413

The {A, C} configurations of the previous numerical anal- 414

ysis were tested individually and the activation power Pminav,R 415

was measured, frequency by frequency, as the minimum power 416

emitted by the reader’s generator so that the chip ID was cor- 417

rectly decoded by the receiver. The transducer power gain is 418

hence computed, from (1) as GT = pC/Pminav,R. Simultaneously, 419

the backscattered power PR←T was recorded and the round 420

trip gain was derived from (2) as GRT = PR←T/Pminav,R to 421

make some considerations on the expected BER (that can 422

not be directly measured with our facilities). Measurements 423

IEEE P

roof

MIOZZI et al.: CONSTRAINED SAFETY-INTEGRITY PERFORMANCE OF THROUGH-THE-ARMS UHF-RFID TRANSCUTANEOUS WIRELESS COMMUNICATION 7

Fig. 9. Comparison among measurements and simulations for tags implantedin position A and C concerning (a) the transducer power gains and (b) theround trip gains. Horizontal lines in the GRT diagrams indicate the minimumvalue that is compliant with BER < 10−3 constraint for a reliable link byassuming a reader power of 23 dBm. Finally, (c) reports the extrapolatedactivation powers for the case of worse sensitivity sensing-oriented chips.Horizontal lines indicate the maximum power that is compliant with the SARconstraint.

are compared with the simulated results obtained from (11)424

of the corresponding experimental arrangements. As the true425

implantation depth in the experimental setup can not be accu-426

rately controlled, simulations were performed for two different427

implantation depths {5 mm, 10 mm}. Overall, there is a428

substantial coherence between measurements and simulations429

(Fig. 9a-b). A frequency shift of about 100 MHz between the430

tag in position A and position C (visible in both simulations431

and measurements) is due to the different coupling between432

tag and reader’s antenna.433

Focusing to the round trip gains, the comparison with the434

minimum value (GRT > −52.5 dB) that is compliant with the435

BER < 10−3 constraint, shows how the interrogation of tag436

implanted in the position C is borderline, independently on the437

chip sensitivity, as it was found in simulations. Instead there438

is a comfortable (20 dB) margin for the tag in position A.439

Starting from above two-port networks gains, the activation440

powers for sensor-oriented microchip (pC = −10 dBm) are441

Fig. 10. Differences in the minimum activation power �Pmin of the mis-alignments reader-tag at 867 MHz w.r.t. the links in {A, C}. Simulated valuesare referred to a noise band of �f = 1MHz as in measurements.

extrapolated in Fig. 9c and have to be compared with the 442

maximum power (23 dBm) that is compliant with SAR limits. 443

A sensor-tag in position A could be easily read while tag in 444

position C would once again result in a borderline condition, 445

at least in this simplified conservative phantom. 446

Finally, Fig. 10 resumes the relative minimum activation 447

power (w.r.t. perfectly aligned configurations) vs. longitudinal 448

and axial misalignments of the slot antenna (keeping the tag 449

fixed). In spite of the challenge in reproducing the simulated 450

tests in Fig. 6, the degradation of Pminav,R due to the misalignment 451

found in the experiment follows the same trend of simulations 452

with the full layered arm model, even if some discrepan- 453

cies with measurements are visible for increasing linear and 454

angular misalignments, especially in the most challenging con- 455

figuration C probably due to the approximated knowledge of 456

the phantom permittivity (the minced meat is dryer than the 457

in-vivo muscle). 458

VI. CONCLUSION 459

Even though the compliance with the regulation on elec- 460

tromagnetic exposure of the human body enforces a severe 461

limitation to the maximum power that can be emitted by 462

the reader, the considered UHF-RFID transcutaneous teleme- 463

try allows establishing a backscattering link with a reliable 464

BER < 10−3. Communication can be achieved with a 465

transponder implanted on the agonist muscle (just below the 466

interrogating antenna) and even on the antagonist muscle in 467

the opposite position, even if at the cost of a narrower power 468

margin. An angular displacement up �φ = ±35 ◦ and lin- 469

ear displacement up to 2.5 cm with respect to the reader’s 470

axis, could be tolerated in the most convenient configuration. 471

Moving from a low-sensitivity chip to a more performing one, 472

the scenario is not radically changed, as the communication 473

with tags out of the reader’s axis would still return a low 474

BER. Multiple reader’s antennas could be used to enlarge the 475

read region inside the stump. In future research, a working 476

IEEE P

roof

8 IEEE JOURNAL OF RADIO FREQUENCY IDENTIFICATION

prototype will be arranged to investigate the signal-level top-477

ics concerning the required sampling rate to recognize some478

muscular patterns and the electromagnetic susceptivity of the479

EMG sensor to the interrogating electromagnetic field.480

REFERENCES481

[1] F. Cordella et al., “Literature review on needs of upper limb prosthesis482

users,” Front. Neurosci., vol. 10, p. 209, May 2016.483

[2] D. Farina and R. Merletti, Surface Electromyography: Physiology,484

Engineering, and Applications, 1st ed. Piscataway, NJ, USA: Wiley,485

2016.486

[3] P. Morel et al., “Long-term decoding of movement force and direction487

with a wireless myoelectric implant,” J. Neural Eng., vol. 13, no. 1,488

pp. 1–14, Feb. 2016.489

[4] F. Riillo et al., “Evaluating the influence of subject-related variables on490

EMG-based hand gesture classification,” in Proc. IEEE Int. Symp. Med.491

Meas. Appl., Lisbon, Portugal, Jun. 2014, pp. 1–5.492

[5] B. D. Farnsworth, D. M. Taylor, R. J. Triolo, and D. J. Young, “Wireless493

implantable EMG sensing microsystem,” in Proc. 7th IEEE Sensors494

Conf., Lecce, Italy, Oct. 2008, pp. 1245–1248.495

[6] B. D. Farnsworth, D. M. Taylor, R. J. Triolo, and D. J. Young,496

“Wireless in vivo EMG sensor for intelligent prosthetic control,” in Proc.497

Solid-State Sensors Actuators Microsyst. Conf., Denver, CO, USA, 2009,498

pp. 358–361.499

[7] D. McDonnall, S. Hiatt, C. Smith, and K. S. Guillory, “Implantable500

multichannel wireless electromyography for prosthesis control,” in501

Proc. 34th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. (EMBC),502

San Diego, CA, USA, 2012, pp. 1350–1353.503

[8] R. F. Weir, P. R. Troyk, G. A. DeMichele, D. A. Kerns, J. F. Schorsch,504

and H. Maas, “Implantable myoelectric sensors (IMESs) for intramus-505

cular electromyogram recording,” IEEE Trans. Biomed. Eng., vol. 56,506

no. 1, pp. 159–171, Jan. 2009.507

[9] G. A. DeMichele, Z. Hu, P. R. Troyk, H. Chen, and R. F. Weir, “Low-508

power polling mode of the next-generation IMES2 implantable wireless509

EMG sensor,” in Proc. 36th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc.510

(EMBC), Chicago, IL, USA, 2014, pp. 3081–3084.511

[10] P. F. Pasquina et al., “First-in-man demonstration of a fully implanted512

myoelectric sensors system to control an advanced electromechanical513

prosthetic hand,” J. Neurosci. Methods, vol. 244, pp. 85–93, Apr. 2015.514

[11] S. Amendola, R. Lodato, S. Manzari, C. Occhiuzzi, and G. Marrocco,515

“RFID technology for IoT-based personal healthcare in smart spaces,”516

IEEE Internet Things J., vol. 1, no. 2, pp. 144–152, Apr. 2014.517

[12] V. Di Cecco, S. Amendola, P. P. Valentini, and G. Marrocco, “Finger-518

augmented RFID system to restore peripheral thermal feeling,” in Proc.519

IEEE RFID Conf., Phoenix, AZ, USA, 2017, pp. 54–60.520

[13] C. Occhiuzzi, G. Contri, and G. Marrocco, “Design of Implanted RFID521

Tags for Passive Sensing of Human Body: The STENTag,” IEEE Trans.522

Antennas Propag., vol. 60, no. 7, pp. 3146–3154, Jul. 2012.523

[14] R. Lodato, V. Lopresto, R. Pinto, and G. Marrocco, “Numerical and524

experimental characterization of through-the-body UHF-RFID links525

for passive tags implanted into human limbs,” IEEE Trans. Antennas526

Propag., vol. 62, no. 10, pp. 5298–5306, Oct. 2014.527

[15] E. Moradi et al., “Backscattering neural tags for wireless brain-machine528

interface systems,” IEEE Trans. Antennas Propag., vol. 63, no. 2,529

pp. 719–726, Feb. 2015.530

[16] C. Miozzi, S. Guido, G. Saggio, E. Gruppioni, and G. Marrocco,AQ4

531

“Feasibility of an RFID-based transcutaneous wireless communication532

for the control of upper-limb myoelectric prosthesis,” in Proc. 12th Eur.533

Conf. Antennas Propag., London, U.K., Apr. 2018.534

[17] G. Saggio, L. Bianchi, S. Castelli, M. B. Santucci, M. Fraziano, and535

A. Desideri, “In vitro analysis of pyrogenicity and cytotoxicity pro-536

files of flex sensors to be used to sense human joint postures,” Sensors,537

vol. 14, no. 7, pp. 11672–11681, 2014.538

[18] R. Merletti and P. Parker, Electromyography: Physiology, Engineering,539

and Noninvasive Applications. Hoboken, NJ, USA: Wiley, 2004.540

[19] A. Sheibani and M. A. Pourmina, “Study and analysis of EMG signal541

and its application in controlling the movement of a prosthetic limb,”542

Health Technol., vol. 6, no. 4, pp. 277–284, Dec. 2016.543

[20] G. Li, Y. Li, L. Yu, and Y. Geng, “Conditioning and sampling issues544

of EMG signals in motion recognition of multifunctional myoelec-545

tric prostheses,” Ann. Biomed. Eng., vol. 39, no. 6, pp. 1779–1787,546

Feb. 2011.547

[21] D. Dobkin, The RF in RFID. Burlington, MA, USA: Elsevier, 2012.548

[22] M. J. Ackerman, “The visible human project,” Proc. IEEE, vol. 86, no. 3, 549

pp. 504–511, Mar. 1998. 550

[23] M. K. Gospal and T. C. Cetas, “Current sheet applicators for clinical 551

microwave hyperthermia,” IEEE Trans. Microw. Theory Techn., vol. 41, 552

no. 3, pp. 431–437, Mar. 1993. 553

[24] S. Manzari and G. Marrocco, “Modeling and applications of a chemical- 554

loaded UHF RFID sensing antenna with tuning capability,” IEEE Trans. 555

Antennas Propag., vol. 62, no. 1, pp. 94–101, Jan. 2014. 556

[25] Loopetto Mecstar Srl. [Online]. Available: http://www.mecstar.it/AQ5

557

it/prodotti/rfid-uhf/loopetto 558

[26] “Limitation of exposure of the general public to electromagnetic fields,” 559

Eur. Council, Strasbourg, France, Eur. Council RecommendationAQ6

560

1999/519/EC, Jul. 1999. [Online]. Available: http://ec.europa. 561

eu/enterprise/electr_equipment/lv/rec519.pdf 562

[27] S. J. Orfanidis. Electromagnetics Waves and Antennas. [Online]. 563

Available: http://www.ece.rutgers.edu/~orfanidi/ewa 564

[28] F. Fuschini, C. Piersanti, F. Paolazzi, and G. Falciasecca, “Analytical 565

approach to the backscattering from UHF RFID transponder,” IEEE 566

Antennas Wireless Propag. Lett. vol. 7, pp. 33–35, 2008. 567

[29] R. E. Collin, Antennas and Radiowave Propagation. New York, NY, 568

USA: McGraw-Hill, 1985. 569

[30] Datasheet AMS-SL900A. [Online]. Available: http://ams.com/ 570

documents/20143/36005/SL900A_DS000294_4–00.pdf 571

[31] RFMicron’s Xerxes. [Online]. Available: http://rfmicron.com/rfm5000- 572

xerxes-wireless-passive-sensor-ic/ 573

[32] FARSENS Rocky 100. [Online]. Available at http://www. 574

farsens.com/products/rocky100 575

[33] M. W. A. Khan, M. Rizwan, L. Sydänheimo, T. Björninen, Y. Rahmat- 576

Samii, and L. Ukkonen, “Characterization of 3-D loop antenna to 577

overcome the impact of small lateral misalignment in wirelessly pow- 578

ered intracranial pressure monitoring system,” IEEE Trans. Antennas 579

Propag., vol. 65, no. 12, pp. 7405–7410, Dec. 2017. 580

[34] Datasheet Monza R6. [Online]. Available: https://support.impinj.com/ 581

hc/en-us/articles/202765328-Monza-R6-Product-Brief-Datasheet. 582

[35] R. Jegadeesan, S. Nag, K. Agarwal, N. V. Thakor, and Y.-X. Guo, 583

“Enabling wireless powering and telemetry for peripheral nerve 584

implants,” IEEE J. Biomed. Health Inform., vol. 19, no. 3, pp. 958–969, 585

May 2015. 586

Carolina Miozzi received the M.Sc. degree in med- 587

ical engineering from the University of Rome Tor 588

Vergata in 2017, where she is currently pursu- 589

ing the Ph.D. degree in electronics engineering. 590

Her research interests include on-skin antennas and 591

through-the-body wireless communication systems 592

for smart health monitoring and detection of bio- 593

physical parameters. She is currently researching 594

on the development of a wireless power transfer 595

and transcutaneous telemetry system for the con- 596

trol of robotic limb prostheses, using EMG-sensors 597

implanted subcutaneously. 598

Giovanni Saggio received the degree in electronics 599

engineering and the Ph.D. degree in microelectron- 600

ics and telecommunication engineering from the 601

University of Rome Tor Vergata, Italy. 602

He is a Researcher and an Aggregate Professor 603

with the University of Rome Tor Vergata, where 604

he holds the chairs about electronics with the 605

Engineering Faculty and the Medical Faculty. He has 606

authored or coauthored about 200 scientific publica- 607

tions, eight patents, several book chapters, and is a 608

unique author of four books about electronics. His 609

research interests involve sensors, biotechnologies, human–machine systems, 610

motion capture, and machine learning. He co-founded the Captiks enterprise 611

involved in systems to measure and analyze human kinematics, the Seeti enter- 612

prise involved in virtual and augmented reality systems, and the VoiceWise 613

enterprise involved in early diagnosis by means of voice analysis. 614

IEEE P

roof

MIOZZI et al.: CONSTRAINED SAFETY-INTEGRITY PERFORMANCE OF THROUGH-THE-ARMS UHF-RFID TRANSCUTANEOUS WIRELESS COMMUNICATION 9

Emanuele Gruppioni was born in Bologna, Italy,615

in 1975. He received the master’s degree in elec-616

tronic engineering (biomedical specialization) from617

the University of Bologna in 2006. During the uni-618

versity studies, he was a Technician and the System619

Administrator for private companies. At the end of620

the studies, he obtained a scholarship in the field621

of organization and innovation in company man-622

agement. Since 2006, he has been a Researcher for623

INAIL Prosthesis Centre. In 2014, he received the624

Degree of Certified Prosthetist and Orthotist, issued625

by the University of Bologna, where he has been a Professor of the course of626

“Orthoses and Orthopedic Aids III,” Faculty of Medicine and Surgery since627

2015. He is currently the Chief of Research Area with INAIL Prosthesis628

Centre, researching as a Project and the Technical Manager in several projects629

in collaboration with the Italian Institute of Technology, Genova, University630

“Campus Bio-Medico,” Rome, Scuola Superiore Sant’Anna, Pisa, National631

Research Council, and Politecnico di Milano. His activities mainly concern632

the design and development of prosthetic limbs and new medical devices for633

rehabilitation. He is a Reviewer of IEEE.634

Gaetano Marrocco received the Laurea degree 635

in electronic engineering and the Ph.D. degree in 636

applied electromagnetics from the University of 637

L’Aquila, Italy, in 1994 and 1998, respectively. 638

He is a Researcher with the University of Rome 639

Tor Vergata from 1994 to 2014, an Associate 640

Professor of electromagnetics from 2013 to 2017, 641

a Guest Professor with the University of Paris-est 642

Marne la Vallèe in 2015, and has been a Full 643

Professor with the University of Roma Tor Vergata 644

since 2018, where he currently serves as the Director 645

of the Medical Engineering School and the Pervasive Electromagnetics Lab. 646

His research is mainly directed to the modeling and design of broadband and 647

ultrawideband antennas and arrays as well as of sensor-oriented miniaturized 648

antennas for biomedical engineering, aeronautics and radiofrequency identi- 649

fication (RFID). During the last decade, he carried out pioneering research 650

on the wireless-activated sensors, and in particular, on epidermal electron- 651

ics, contributing to move from the labeling of objects to the passive sensor 652

networks in the Internet of Things era. 653

Prof. Marrocco is the Co-Founder and the President of the University 654

spin-off RADIO6ENSE active in the short-range electromagnetic sensing 655

for Industrial Internet of Things, smart manufacturing, and automotive and 656

physical security. He serves as an Associate Editor for the IEEE JOURNAL 657

OF RADIOFREQUENCY IDENTIFICATION. He is the Chair of the Italian 658

Delegation URSI Commission D Electronics and Photonics. He was the Chair 659

of the Local Committee of the V European Conference on Antennas and 660

Propagation in 2011, the TPC Chair of the 2012 IEEE-RFID TA in Nice, 661

France, and the TPC Track-Chair of the 2016 IEEE Antennas and Propagation 662

International Symposium and the IEEE-RFID 2018 USA. He was a member 663

of the IEEE Antennas and Propagation Society Awards Committee. 664